AI: An Artist’s Friend or Foe? | Dan Jeffries, CIO of Stability.ai (Stable Diffusion)

Show Notes & Transcript:

🚀 After Effects Kickstart 🚀

[joeykorenman]

This comments section is everything I dreamed it would be.

[saidasaetgar]

1) Every AI tool vs replacement convo should start with definitions: Narrow AI is what we have today (MJ, SD, Chat-GPT) it performs very specific tasks like image generation. Artificial General Intelligence (AGI) refers to an agent or an interconnected network of narrow AI's capable of performing any intellectual task, including coming up with extremely creative solutions previously unimaginable. AGI is on its way to becoming a reality but it’s not here yet.

Folks who think AI will stay narrow in the future (i.e. won't change much from it's current state) are more likely to see MJ/SD/chat-GPT as a tool

Folks who think AGI will pop up in the next 10 years tend to think of it as a replacement rather than tool

2) We can’t separate AI technology from the context it’s arriving into i.e. profit-oriented capitalism. The speed of AI art development this year has been insane (remember how MJ stuff used to look like back in May?). We aren’t ready for AI that improves so fast, especially since it's being adopted and iterated in a world of shareholder-centered business models and unaligned collective values

3) We shape our tools and then the tools shape us back. If natural selection could be applied to AI, what art / behaviors / cultures would AI select for? Are these values human-positive or profit-hype-centric?

4) It feels somewhat unfair that AI is on its way to automate creative art jobs instead of boring manual tasks that nobody wants to do or nobody should be doing because they are dangerous (e.g. labour jobs that expose workers to toxic chemicals). This situation reminds me of the book "Waste Tide" by Chen Qiufan, a chinese cyberpunk thriller where, despite technology's progress, e-waste recycling is still done by humans in sweatshops because it's cheaper than coming up with tech solutions for it. It might be an overreaction to compare our current situation to that. BUT What I'm trying to say is that I hope AI advances in 2023 will focus on more world-positive utilitarian pursuits rather than further undermining the value of human expression and creativity.

[andreasgaschka]

49:00 AI is not learning like humans, and this guy knows better. He's telling this lie to protect his company. Machine Learning is all about data compression: You train a model by compressing data into it. If you want to use my data to train your model you get rich with, get permission and then pay me royalties. No one wants to copyright a style, but I do have copyright on my IPs and art. It's not about Opt-Out, it's about Opt-In and when you learn how these companies operate to gather their training data you immediately realize, they know that they act illegal or at least immoral.

[fy]

Guns: Friend or Foe? We speak to the National Rifle Association president Charles Cotton.

[jeekim]

I really wanted this conversation to resolve my qualms about these AI tool - or, more specifically, the ENTITLEMENT of these tool creators that have obscenely naive optimism toward its benefits while willfully turning a blind eye to the integrity of the creative process. I feel like while Joey carefully articulates the concerns of the artists, Daniel superficially gestures sympathy but in reality, answers with the singular attitude of "civilizations evolve with technology - get over it."

I think what the creative community wants/needs to hear is accountability from these tool creators. Setting guardrails, parameters for what these tools are, and what they are NOT. As long as their objective is to attract/maximize users of their tool, it is highly likely "what makes art, art" is the least of their concerns. Of course, they can afford to feign fake sympathy because whatever economic, existential cost these tools bring - it's not for them to pay.

Please note, I'm not AT ALL saying these tools are evil. These are EXCELLENT tools. If anything, as somebody who had to make a bunch of pitch decks and fill them up with very specific visual references for a job a few years back, I wish these tools were available then. Even now, if you're an artist, you're always looking for inspiration through references. And Daniel is right - prototypes using AI can definitely expedite and even enrich the creative process. So, yes, he is absolutely right in terms of preaching the infinite possibilities these tools offer - but I think he missed the purpose of this platform, which was to meet somewhere in between the hyper-anxiety of the creative community and the hyper-optimism of the tech community.

[g.u.d.-graciousupondeath]

While AI is something to discuss, I don't think giving Daniel a platform to speak and defend a tool designed to steal from artists is a healthy decision. To put it bluntly, it's weird that a service about educating designers is highlighting and offering a platform to someone and a service designed to hurt designers.

[marcusv]

AI might assist our jobs for now but soon it will take over. It's naive to think it's here to help and all we have to do is adapt and embrace. Going to make it harder and harder to find employment and be of value as a human to a company and the world.

[misanthropiclusion]

only reason for it's existance is to cheapen labor and the CEO has had quite a disgusting and incredibly unethical view on his personal social media, he needs a lesson

[GalaxColor]

Already wrong approach when you know for a fact that Emad clearly said he wanted to destroy the illustration industry 🙁.

[GalaxColor]

As an artist, you are learning by copying from experts, but you are not saying it's your work. The work you actually have originally is the art style you adapt from experiences. To say ai is similar to a conscious meaning is just plain stupid. Artists can take inspiration from other artists but they
will put their own personal preferences into the
artwork. Ai cannot recreate a personal touch because
it's only relying on the art fed to it. Artists will learn art
from scratch starting from "learning artist to
professional artist" and along this journey creating
familiar brush strokes patterns and certain ways of
portraying anatomy. While they can certainly be
inspired by another artist they will fall into a category
of their own becoming a new style along that journey
Ai can never do this; it can never have a journey of
learning from scratch and creating its own unique
way of painting, because all it`s doing is using up
artwork from its datasets. This is why every piece of
Ai "art" prompt you see can be linked to an existing
art style.

[becausejamie]

Tech wants to dehumanize the process of creativity and art...why, money. Corporations no longer want to pay people for time as shareholders and costs place pressure on margins. I will always root against AI and it's formalization into the world of art. This conversations only goal is to normalize AI THEFT and I'm resisting the pull to digress into a rant about the lack of empowering uncreative. I avoid division, but here is the line in the sand....Us verse them is happening and growing.

[GalaxColor]

This is not a new media. It feeds upon millions of DIGITAL work, digital art with WORK from humans. I seriously do not care if the CIO is an artist, he has no right to speak for everyone. Having too much pride in this software is not okay. Image AI is not a camera, nor akin to any other invention. The practical similarities are extremely meager in any case. AI brings with it a slew of unique social and ethical concerns that the camera and prior inventions did not. This type of argument is simply reductive. There may be a loose similarity to previous events (the camera, the printing press etc.) but the implications of the technology and the way it's built are entirely different. Photographers and digital artists are very different, even if they work digitally. It's THEIR WORK. Ai has no work done by the person who inputs the prompt. Artists are not people who put prompts in. That is seriously just an idea. As another artists who's work has been stolen for AI use I have all of the right to be mad at someone who refuses to see the impact this has on others. Absolutely disgusting

[GalaxColor]

No one said anything about copyrighting styles. NO ONE. We want PERMISSION. Is it really that hard to think about?? It's really as simple as asking tour friend for the homework. "Hey, can I copy your homework?" YOU ASK FOR PERMISSION. You are in the wrong here for not doing so.

[RidvanMaloku]

The current AI models are not here to help the artists, they are to take from artists without any consent!

[mr.person]

Why not make a class action lawsuit if the company is in America? This is stealing, not copying. The CEO seems to be someone that doesn't understand ethics or hard work put into an artists' work. Yes, it took let's say 1, 000 hours total to create the code, but it stole millions of hours from the creatives in the industry.

[ChitlinsLaundry]

Fascinating stuff, thank you both for open the conversation !

[Mechamedia_Industries]

Great conversation, but the numerous jump cuts are very distracting and make me wonder what was edited out.

[esgarlara]

Tremendous talk! The problem I see is that we are not responsible enough to use AI for what it is, a tool... Only the future can prove us right or not.

[robertmunciework]

My take is that there is a lot of hype around generative AI because that's the intent of the people who made it.
Anyone who knows a bit about neural networks knows that you need to stress test it so that it becomes more useful.

It needs more input (not just art in this case) but prompt information from people. The AI needs to learn from its successful outputs to build more meaningful connections. Every time we use OpenAi’s apps we in theory are contributing to the programs ability to successfully turn text into images.

Why I don't think AI is coming for anyones job (at least anytime soon) is it seems there are hard limits to its flexibility, repeatability and scalability.

Anyone who's worked for any design company knows these three things need to be accounted for before you adopt any new technology; especially a technology that’s intent is to reduce labor costs.

AI art fails at every single one of these issues.

Flexibility:

The same prompt delivered by two different designers, at two different computers, produce two widely different outputs. The output is randomised to a certain extent which yields interesting outputs but not useful if you have to make iterations on the content in the same style. Some people have actually gotten around this limitation by using some of these Ai’s basic iteration functions and editing tools but this proves the point that a human hand needs to guide these generative images from start to finish. When a designer spends a long time fussing with prompts, outputs, the AI’s clunky editing tools and sometimes even Photoshopping outputs. This often results in a situation where a talented artist and a person prompting could take the same amount of time producing a deliverable.

Repeatability:

Lets say though you have an output that is right on the money, a perfect generative piece that needs very little tweaking. There is no guarantee your next output is going to be of that quality and style. There is unfortunately a lot left to random chance with these systems. The main appeal of Ai, typing in a prompt and wondering what you are going to get is exactly its draw back. People who make images for a living have gotten use to specific software or tool workflows. Years of experience tell them exactly what they need to do in what order to get the exact result they want. A good designer has a pretty good image in their head on what output they want and can quickly reproduce that image with tools. Often times converting that image in their head into text and seeing what an AI produces results in an accurate recreation of words, but often times is worlds away from the actual output the designer wants.

Scalability:

You’ll need to make your own neural network for your company’s uses. Eventually you will generate so many images you’ll want to keep your brand consistency. Here’s the problem you need millions of images to train an Ai so that it’s even remotely useful. Which means you will have to spend a lot of time on another company’s Ai feeding it your company’s intellectual property until you are even remotely close to having a large enough dataset to go in-house with your own generative AI.

Cryptocurrency, NFTs and AI, it’s all just a Silicon Valley hype train because they are desperate for funding; they know eventually Disney and Viacom are going to find out their content was used to train these AIs and their lawyers are coming.

PS: Remember a decade ago they promised we'd all have self driving cars. Now self driving cars are a literal joke.

[harismirza]

"One day the AIs are going to look back on us the same way we look at fossil skeletons on the plains of Africa."